np NP, Friday September 30, 2016

[Vorlonagent 442]

11 minutes ago, hkmaly said:

Oooooh ... imbedded programing. Never done any. Must be fun....

Not really.  I had no idea how to do embedded code.  I was working for an electrical contractor who wanted to branch out into software.  Unfortunately, the owner made promises to the customer out of ignorance of software development times.  He had limited code in place (it could give collect and transmit temperature data) and promised a complete system in 3 months.  He had first hired a high school kid (son of a computer sci university prof, but still high school) then let him go and got a friend of mine.  He needed help and that's how I got the job.  My first day on the job, the project was already a year late.  The customer was not happy with the situation but they never took it out on us.

We did a lot of the work at the customer's business in order to show them that we were indeed working on their stuff, but we were flying largely blind, either writing our own code or trying to salvage the code of the kid who came ahead of us.

Eventually the customer started withholding money from other work and the whole project abruptly collapsed along with the job one Friday afternoon.  We were told to take everything of value we could carry by end of day and leave.  We were supposed to start working back at the normal office but by monday I was unemployed.  I'm pretty sure the software project collapsed the entire company.

[hkmaly 490]

18 minutes ago, Vorlonagent said:

We were told to take everything of value we could carry by end of day and leave.

Ooooh. The classical "Before we start evaluating the state of company, I would recommend to have a snack, while we still have any".

That's even more interesting story than I expected. Unfortunately it ended badly, unless you managed to grab something valuable that Friday.

[mlooney 3,417]

19 hours ago, Vorlonagent said:

As a DOS user I've seen many EMM386 errors

EMM386 is freaking evil.  Back in the day (mid 80's) I was writing code that through the magic of DOS/16M (and later DOS 4/G when the 386 came out) libraries than ran in protected mode on top of MS-DOS 3.3 (and later DOS 6.22).  Great fun that was.  That's where I picked up my fetish that pointers in c are not an integer type, and treating them like one will result in disaster.  Which is why my project never got ported to c++, where NULL =(void *)0 was hard coded into the language.  Also great fun when you had to deal with a memory map (real mode) device from protected mode.  In 286 protected mode, even looking past the memory assign to a pointer by malloc() would result in a crash, never mind trying to write past one.   Lint became my friend.  In fact our "add to the version control" batch file would not let you do so until lint returned a 0, and damn it was strict.

 

[Vorlonagent 442]

15 hours ago, hkmaly said:

Ooooh. The classical "Before we start evaluating the state of company, I would recommend to have a snack, while we still have any".

That's even more interesting story than I expected. Unfortunately it ended badly, unless you managed to grab something valuable that Friday.

No, the company kept some stuff.  I was too honest.   Mostly we brought home a couple of 8086 PCs and some of the relays we were using for compressor control.

The owner was already looking at a deal that involved new (by late-1980s standards) food packaging technology and a startup facility for cranking packaged food out.  I suspect MLooney and others who ate US Army MREs might have my ex-boss at least somewhat to blame for the taste of the food. 

The big thing I got out of it was PC Magazine and an enduring love of pizza.  When lunchtime came I went over to a local pizza place and either went over my code or read PC Magazine.  Between what I picked up and what the other programmer knew we were able to take a 286 motherboard sitting around doing nothing and make a system out of it.  Its 8Mhz processor blew away our 6 MHZ 8086 machines. 

By then I knew there was this company called AMD out there that Intel licensed to help with 286 demand (Intel literally couldn't make them fast enough).  Intel froze them out of the 386 so they sold budget 286 chips overclocked past intel's self-imposed upper limit of 12MHZ.  16 MHZ or even (gasp!) 20MHZ.  Intel once again went back to the "brain damaged chip you can afford" well and created the 386SX.  32-bit processor, 16-bit data path. 

The first computer I ever bought was a 386SX.  It was also the most expensive computer I ever bought.  I think it was over $3000 for a 16MHZ 386SX, 40MB hard drive, 1 MB of memory and Windows 3.0.  Win 3.1 came out only a few weeks later.  I was so upset...  It served me well.  By the time I was ready to upgrade, I was building my own machines.  I developed a price/performance rule that mostly still serves me today.  "The part you want costs $100 - $125, no matter what part we're talking about."  It held for motherboards, memory, drives, and AMD processors.  I've had to modify some of the amounts over the years.  Not everything still costs that much but there's usually an amount you can expect to pay that stays stable over the years until something really big changes.  My memory estimate has dropped to $70, for example.

[hkmaly 490]

7 hours ago, mlooney said:

On 10/06/2016 at 8:50 PM, Vorlonagent said:

As a DOS user I've seen many EMM386 errors

EMM386 is freaking evil.  Back in the day (mid 80's) I was writing code that through the magic of DOS/16M (and later DOS 4/G when the 386 came out) libraries than ran in protected mode on top of MS-DOS 3.3 (and later DOS 6.22).  Great fun that was.

That was based on DPMI, right? I wouldn't blame EMM386 for that. First, it WAS a separate driver. Second, even just specification for that was horrible. You should've used something which will allow you to ring 0.

8 hours ago, mlooney said:

Also great fun when you had to deal with a memory map (real mode) device from protected mode.  In 286 protected mode, even looking past the memory assign to a pointer by malloc() would result in a crash, never mind trying to write past one.

Difference between good and bad design is that in bad design, application under specific extender will crash on incorrect memory access. In good design, you can write program using standard POSIX stuff like signal and mmap, install signal handler on SIGSEGV, handle the problem with mmap and return and the application will continue to run.

(Note that while signal and mmap are POSIX, I don't think this behaviour is guaranteed by POSIX ...)

4 hours ago, Vorlonagent said:

By then I knew there was this company called AMD out there that Intel licensed to help with 286 demand (Intel literally couldn't make them fast enough).

While possible there was problem with demand, note that IBM directly told Intel that there will be no deal (for IBM PC) if they would be only one doing the chips. Intel did cross-licensing agreement with AMD and 80286 was under this agreement (also 8086 and 80186 of course). Intel claimed 80386 was not and until AMD won the case Intel had monopoly on them.

4 hours ago, Vorlonagent said:

The first computer I ever bought was a 386SX.

The first computer I could maybe call mine was also 386SX. But parents bought it, no idea for how much. 20MHz, 20MB HDD.

 

[Vorlonagent 442]

6 minutes ago, hkmaly said:

While possible there was problem with demand, note that IBM directly told Intel that there will be no deal (for IBM PC) if they would be only one doing the chips. Intel did cross-licensing agreement with AMD and 80286 was under this agreement (also 8086 and 80186 of course). Intel claimed 80386 was not and until AMD won the case Intel had monopoly on them.

I hadn't heard that.  I can see how the lawsuit went:  Intel sez "IBM didn't make a computer on this chip so we're not bound by the agreement."  AMD sez "Oh yes you are" and won.  In the meantime, Intel builds the 386SX to try to attack AMD's 286 chip sales and maybe shut the company down.

[hkmaly 490]

17 minutes ago, Vorlonagent said:

28 minutes ago, hkmaly said:

While possible there was problem with demand, note that IBM directly told Intel that there will be no deal (for IBM PC) if they would be only one doing the chips. Intel did cross-licensing agreement with AMD and 80286 was under this agreement (also 8086 and 80186 of course). Intel claimed 80386 was not and until AMD won the case Intel had monopoly on them.

I hadn't heard that.  I can see how the lawsuit went:  Intel sez "IBM didn't make a computer on this chip so we're not bound by the agreement."  AMD sez "Oh yes you are" and won.  In the meantime, Intel builds the 386SX to try to attack AMD's 286 chip sales and maybe shut the company down.

I didn't see it in news back then either. You can read it on wikipedia. Note that it took years before AMD won.

[Scotty 1,061]

21 minutes ago, Vorlonagent said:

32 minutes ago, hkmaly said:

While possible there was problem with demand, note that IBM directly told Intel that there will be no deal (for IBM PC) if they would be only one doing the chips. Intel did cross-licensing agreement with AMD and 80286 was under this agreement (also 8086 and 80186 of course). Intel claimed 80386 was not and until AMD won the case Intel had monopoly on them.

I hadn't heard that.  I can see how the lawsuit went:  Intel sez "IBM didn't make a computer on this chip so we're not bound by the agreement."  AMD sez "Oh yes you are" and won.  In the meantime, Intel builds the 386SX to try to attack AMD's 286 chip sales and maybe shut the company down.

Was that the same court case that forced Intel to stop trademarking the x86 as a brand and start going with Pentium, Pentium Pro and Xeon? Of course AMD ended up going with the K5, K6 and Athlon

[hkmaly 490]

2 minutes ago, Scotty said:

Was that the same court case that forced Intel to stop trademarking the x86 as a brand and start going with Pentium, Pentium Pro and Xeon? Of course AMD ended up going with the K5, K6 and Athlon

IMHO the patent office declaring numbers non-patentable likely wasn't related, at least not officially.

[ProfessorTomoe 1,308]

Most of my CPUs since my original 80286-12 have been AMD, although I briefly had a NexGen 586 at one point while playing hardware guinea pig for 3D Realms. Only lately have I begun to make the shift back to Intel, most recently with the i7-6700HQ in my Gigabyte laptop.

Remember upgrading RAM on the old 286s? I had a motherboard that I upgraded to capacity, one chip at a time. Even had a special chip-puller tool on stand-by in case I bent the legs on any of them.

[lonjil 11]

I forgot to quote it, but a few pages ago people were talking about IBM making really dumb mistakes, and I just wanted to point that with the amount of business IBM does, stuff like that is bound to happen just though pure chance, and IBM still got through fine. It's not like Xerox or Kodak where their mistakes caused their downfalls.

[The Old Hack 2,448]

19 minutes ago, lonjil said:

I forgot to quote it, but a few pages ago people were talking about IBM making really dumb mistakes, and I just wanted to point that with the amount of business IBM does, stuff like that is bound to happen just though pure chance, and IBM still got through fine. It's not like Xerox or Kodak where their mistakes caused their downfalls.

Fair enough, and it isn't as if any of us really wish the company any ill, I think. We are just marveling at how things happen in the business world.

[lonjil 11]

1 hour ago, The Old Hack said:

Fair enough, and it isn't as if any of us really wish the company any ill, I think. We are just marveling at how things happen in the business world.

Fair enough to you too, I guess :-P

[lonjil 11]

I wanted to put this in an edit, but the edit box seems messed up.

---

The case of Xerox is actually pretty funny, Xerox's Palo Alto Research Center created all kinds of cool, ahead of the times stuff, but because executives didn't really understand why anyone would use that stuff, a lot of it was more or less given away, for example, what became the modern desktop under Apple. I have code running on my computer that is copyright Xerox Corp 1985, because they gave it out for free, under an open license.

Edit: apparently, I'm getting white text on white background in the editing box. Also, does anyone know to make a horizontal bar in the formatting?

Edited October 9, 2016 by lonjil

[mlooney 3,417]

2 hours ago, lonjil said:

Edit: apparently, I'm getting white text on white background in the editing box. Also, does anyone know to make a horizontal bar in the formatting?

Yes, it does that.  Several cures, but the one that also is for your second question.  Open up the "source" option.  That makes every thing visible again, granted in HTML format.  For putting a horizontal bar in use <hr> while in source mode.  Other than that, I don't think there is a way.

[Don Edwards 548]

The easiest cure for the white text is to click on where text ought to be (NOT in a quote box - if you click there then this only works on the contents of that quote box), hit Ctrl-A, and then change the text color (that's the underlined "A" in the row of tool buttons) from "Automatic" to, well, pretty much anything other than white. I go with the black square at the left of the first full row.

This has been complained about, both in the forum and in feedback to the admins, several times.

[hkmaly 490]

On 10/08/2016 at 6:53 AM, ProfessorTomoe said:

Most of my CPUs since my original 80286-12 have been AMD, although I briefly had a NexGen 586 at one point while playing hardware guinea pig for 3D Realms. Only lately have I begun to make the shift back to Intel, most recently with the i7-6700HQ in my Gigabyte laptop.

All my desktop CPUs have been AMD except that first. All my notebook CPUs have been Intel. Although given the number of notebooks I had, it might've been just a coincidence ... the desktop is not coincidence. Intel doesn't have middle-end chips. Everything it's either low-end (judged by power) or high-end (judged by price).

[HarJIT 97]

5 hours ago, Don Edwards said:

The easiest cure for the white text is to click on where text ought to be (NOT in a quote box - if you click there then this only works on the contents of that quote box), hit Ctrl-A, and then change the text color (that's the underlined "A" in the row of tool buttons) from "Automatic" to, well, pretty much anything other than white. I go with the black square at the left of the first full row.

This has been complained about, both in the forum and in feedback to the admins, several times.

I will take the opportunity to link to https://sites.google.com/site/harjitegs/910-invisible-text-problem again (hey, that's what I put in up for!)

[Cpt. Obvious 73]

On den 8 oktober 2016 at 6:53 AM, ProfessorTomoe said:

Most of my CPUs since my original 80286-12 have been AMD, although I briefly had a NexGen 586 at one point while playing hardware guinea pig for 3D Realms. Only lately have I begun to make the shift back to Intel, most recently with the i7-6700HQ in my Gigabyte laptop.

Remember upgrading RAM on the old 286s? I had a motherboard that I upgraded to capacity, one chip at a time. Even had a special chip-puller tool on stand-by in case I bent the legs on any of them.

Ahh... The NexGen 586! That was one oddball CPU, they tried a very unique design that initially didn't work out as well as they hoped for. Basically the CPU core had more in common with RISC designs rather than the CISC design of every other x86 compatible processor. It couldn't even run x86 code natively, but had a chunk of chisel dedicated to translating it to something the core could digest. Add that this was, if not the first then at least one of the first x86 compatibles that wasn't a drop in replacement for an existing intel processor. and the choice of motherboard became severely limited.

But though initial performance hadn't been what man had hoped for they were onto something. Eventually AMD bought them and used some of the IP from NexGen when designing the K6 which was a very successful product.
 

[mlooney 3,417]

Let's see.  My chip sets

1. Digital Equipment Corporation PDP-11 (Not actually mine, but I worked on it for 3 years, starting as a student then got a job a Jr. SysAdmin.  That was in the Ford and Carter years)
2. Cosmic Elf 1802 I had the "expanded" version.  I had 1024 bytes of memory.
3. Nike Hercules Fire Control Computer.  Had vacuum tubes and just enough digital bits to count in this context.  Also not really mine, just one of my Army jobs.
4. Zilog Z80
5. MOS Technology 6502 (actually 6510, same thing but could do weird things with memory paging)
6. Patriot Fire Control Computer.  Again, Army stuff.
7. PDP-11 Again.  Army also.
8. Intel 8088 (Real IBM PC)
9. Intel 80286 (Real IBM AT, which in my arrogant opinion, had the best keyboard ever used on a consumer class computer)
10. A large number various 80386 machines.
11. Intel Pentium Pro (was an home entertainment system with a cable ready video card thing.  Had drivers for Window 95 and that was it.)
12. 3 I don't remember, but were take homes from work any way.
13. PPC G4 (Power PC) which despite the name wasn't in a PC, but an Apple
14. PPC G3
15. PPC G5
16. 2 x PPC G4 (Yes, I bounced up and down the power scale)
17. 3 x Intel Core 2 Duo of various speeds

I never had any thing based on the 68000 series of chips.  I was, and still am, rather attached to a command line option, bash¹ if I can get it.  Apples pre OS X didn't have that option, so I never got a Mac pre OS X. I'm not counting tablets², PDA, phones, etc. in this list.  I've been on Mac's almost exclusively from 2002 on, other than when I had a take home machine.

¹If for some reason I'm on a PC machine that I can't load Linux on, I load cygwin, which give me a bash shell for the command line, and most of the GNU tool chain.

² See all my rants about getting a either Linux or something like that on my RCA table

[Scotty 1,061]

5 hours ago, Cpt. Obvious said:

Ahh... The NexGen 586! That was one oddball CPU, they tried a very unique design that initially didn't work out as well as they hoped for. Basically the CPU core had more in common with RISC designs rather than the CISC design of every other x86 compatible processor. It couldn't even run x86 code natively, but had a chunk of chisel dedicated to translating it to something the core could digest. Add that this was, if not the first then at least one of the first x86 compatibles that wasn't a drop in replacement for an existing intel processor. and the choice of motherboard became severely limited.

But though initial performance hadn't been what man had hoped for they were onto something. Eventually AMD bought them and used some of the IP from NexGen when designing the K6 which was a very successful product.
 

Don't forget the Cyrix 5x85 CPU, that tried to give pentium performance while still using the same socket 3 motherboards that 486's used. AMD tried to market their own 586 for socket 3 but it was more like a souped up 486 whereas Cyrix's cpu more resembled a pentium.

[hkmaly 490]

9 hours ago, Cpt. Obvious said:

Ahh... The NexGen 586! That was one oddball CPU, they tried a very unique design that initially didn't work out as well as they hoped for. Basically the CPU core had more in common with RISC designs rather than the CISC design of every other x86 compatible processor. It couldn't even run x86 code natively, but had a chunk of chisel dedicated to translating it to something the core could digest. Add that this was, if not the first then at least one of the first x86 compatibles that wasn't a drop in replacement for an existing intel processor. and the choice of motherboard became severely limited.

But though initial performance hadn't been what man had hoped for they were onto something. Eventually AMD bought them and used some of the IP from NexGen when designing the K6 which was a very successful product.

That "very unique" design was first time someone tried it, but Pentium Pro, AMD-K5 and ALL x86 CPU after them are based on this idea. AMD-K5 was for example based on AMD Am29000, which was RISC. True CISC CPU would never get anywhere near the performance x86 is now. (Although, technically, current CPUs are not RISC either ... but they still have internal instruction set they translate x86 code to instead of executing it directly.)